Research map / Detection and segmentation / CNN features
End-to-End Trainable Deep Active Contour Models for Automated Image Segmentation: Delineating Buildings in Aerial Imagery
Ali Hatamizadeh, Debleena Sengupta, Demetri Terzopoulos
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add per-pixel learned lambda parameter maps (vs constant lambda) (ablation) | TDAC · Vaihingen (single-instance) | mIoU: 83.79 -> 89.16 | Improved |
| add per-pixel learned lambda parameter maps (vs constant lambda) (ablation) | TDAC · Bing Huts (single-instance) | Boundary F: 48.25 -> 53.50 | Improved |
| add per-pixel learned lambda parameter maps (vs constant lambda) (ablation) | TDAC · Vaihingen (multiple-instance) | mIoU: 83.30 -> 91.10 | Improved |
| add per-pixel learned lambda parameter maps (vs constant lambda) (ablation) | TDAC · Bing Huts (multiple-instance) | mIoU: 68.34 -> 71.30 | Improved |
| add end-to-end trained constant-lambda ACM (Chan-Vese) on top of the CNN backbone | FCN (Ours backbone) · Vaihingen (single-instance) | mIoU: 81.10 -> 83.79 | Improved |
| add end-to-end trained constant-lambda ACM (Chan-Vese) on top of the CNN backbone | FCN (Ours backbone) · Bing Huts (multiple-instance) | mIoU: 60.31 -> 68.34 | Improved |
| add end-to-end trained constant-lambda ACM (Chan-Vese) on top of the CNN backbone | FCN (Ours backbone) · Bing Huts (single-instance) | mIoU: 74.09 -> 73.02 | Mixed |
| replace DarNet with TDAC (CNN+ACM jointly trained) | DarNet · Vaihingen (single-instance) | mIoU: 88.20 -> 89.16 | Improved |
| replace DarNet with TDAC (CNN+ACM jointly trained) | DarNet · Bing Huts (single-instance) | mIoU: 75.20 -> 80.39 | Improved |
| replace Mask R-CNN backbone FCN with TDAC | FCN (Mask R-CNN) · Vaihingen (multiple-instance) | mIoU: 79.42 -> 91.10 | Improved |
| vary convolutional filter size (too small or too large) (ablation) | TDAC · Vaihingen and Bing Huts | mIoU | Mixed |
| increase number of ACM iterations L beyond a threshold (ablation) | TDAC · Vaihingen and Bing Huts | mIoU | No clear effect |
Closest papers by meaning
- Learning Deep Structured Active Contours End-to-End (CVPR 2018 · similarity 0.92)
- DARNet: Deep Active Ray Network for Building Segmentation (CVPR 2019 · similarity 0.89)
- Learning Building Extraction in Aerial Scenes with Convolutional Networks (TPAMI 2017 · similarity 0.88)
- Aerial Lifting: Neural Urban Semantic and Building Instance Lifting from Aerial Imagery (CVPR 2024 · similarity 0.85)
- Boundary-Aware 3D Building Reconstruction From a Single Overhead Image (CVPR 2020 · similarity 0.85)
- SatSynth: Augmenting Image-Mask Pairs through Diffusion Models for Aerial Semantic Segmentation (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
